[Qemu-devel] Using QEMU as an emulator for an industrial machine.

[Qemu-devel] Using QEMU as an emulator for an industrial machine.

[Florent Defay]

Hi,

I need your help:

I am working for an industrial company. Here, we build a machine composed of
- a board with a PowerPC as CPU
- classic devices such as ethernet
- other private devices.

LynxOS runs on the board with private applications and private
additional drivers.

We would like to use QEMU to emulate the whole hardware system so that
APP+KERNEL+DRIVERS run on QEMU. The goal is to debug without
downloading software onto the physical board, but directly on a Linux
PC.
I suppose we'll need to modify QEMU's source to add code to simulate
our private devices.

I would like to know if it is feasible? And then the degree of
difficulty to carry out the new emulator. Can a QEMUL newbie do that?
How much time would be needed? I guess it is very difficult but I need
concrete answers. Any advice is welcome.

Thank you.

Regards,
Florent Defay.

Re: [Qemu-devel] Using QEMU as an emulator for an industrial machine.

[Stefan Weil]

Florent Defay schrieb:
> Hi,
>
> I need your help:
>
> I am working for an industrial company. Here, we build a machine
> composed of
> - a board with a PowerPC as CPU
> - classic devices such as ethernet
> - other private devices.
>
> LynxOS runs on the board with private applications and private
> additional drivers.
>
> We would like to use QEMU to emulate the whole hardware system so that
> APP+KERNEL+DRIVERS run on QEMU. The goal is to debug without
> downloading software onto the physical board, but directly on a Linux
> PC.
> I suppose we'll need to modify QEMU's source to add code to simulate
> our private devices.
>
> I would like to know if it is feasible? And then the degree of
> difficulty to carry out the new emulator. Can a QEMUL newbie do that?
> How much time would be needed? I guess it is very difficult but I need
> concrete answers. Any advice is welcome.
>
> Thank you.
>
> Regards,
> Florent Defay.

Hi,

It is possible, at least up to a certain degree:
You won't be able to emulate time critical applications, for example.

To estimate the amount of time needed, more information is needed:

* Do you have source code for LynxOS? This helps very much to debug
  unexpected behaviour of the emulation.

* Which components are already supported by QEMU (PowerPC variant,
  ethernet, ...), or do you need modifications of existing QEMU code?

* How complex are the unsupported devices (ethernet, private devices)?
  Are these devices well documented? Do they use standard interfaces
  (PCI, USB, memory mapped, ...)?

Three years ago, I tried something like that for an embedded PC board.
Very simple private devices needed a day. Ethernet was missing, so
I wrote hw/eepro100.c which is still unfinished. The most difficult
part here was Windows support (closed source OS!).
You can see the history in qemu-devel or better at
http://repo.or.cz/w/qemu/ar7.git.

Every time when you change your emulation environment (new application,
new version of operating system, change of boot loader, QEMU update, ...),
the emulation can fail because it is never complete.

When you have done the job, you no longer are a QEMU newbie :-)

Regards,
Stefan Weil